Tabulating agreement matrix?

[richart]

Hello,

I have approximately 20 subjects and 30 questions (True or False responses). My goal is to tabulate a respondent-by-respondent matrix, in which each cell would have the number of identical responses per individual. Something like this:

While right now I have something like this, with respondents as rows and questions as columns:

I was going to try and use if-then statements to calculate the agreeance between each combination of respondent, by transposing the data and doing something like "if var1=var2 then agreeance_12=1; else aggreance_12=0" but this would obviously take a very long time. I would appreciate anyone's help!  

[PGStats · Accepted solution]

proc distance gets you close ...

data have;
input subj $ (res1-res5) (:$1.);
datalines;
1 T F T T F 
2 T F F F T
3 F T F T T 
4 F F F F T
;

proc distance data=have out=want method=match shape=square;
id subj;
var nominal (res1-res5);
run;

proc print data=want noobs; run;

                        subj     _1     _2     _3     _4

                         1      1.0    0.4    0.2    0.2
                         2      0.4    1.0    0.4    0.8
                         3      0.2    0.4    1.0    0.6
                         4      0.2    0.8    0.6    1.0

[Ksharp]

data have;
input subj $ (res1-res5) (:$1.);
datalines;
1 T F T T F 
2 T F F F T
3 F T F T T 
4 F F F F T
;
data temp;
 set have;
 array x{*} res:;
 do i=1 to dim(x);
   do j=i+1 to dim(x);
    if x{i}=x{j} then do;a=vname(x{i});b=vname(x{j});output; end;
   end;
 end;
keep a b;
run;
proc freq data=temp noprint;
table a*b/out=temp1;
run;
data temp2(index=(a));
 set temp1;
 output;
 temp=a;a=b;b=temp;output;
 drop temp;
run;

proc transpose data=temp2 out=temp3(drop=_:);
by a;
id b;
var count;
run;
proc sql noprint;
select a into : list separated by ' ' from temp3;
quit;
data temp4;
 retain a &list;
 set temp3;
run;
proc stdize data=temp4 out=want missing=0 reponly;
run;

Assuming I understand your question .

[richart]

This works perfectly as well. Thanks!

[ballardw]

---

@richart wrote:

Hello,

I have approximately 20 subjects and 30 questions (True or False responses). My goal is to tabulate a respondent-by-respondent matrix, in which each cell would have the number of identical responses per individual. Something like this:

!  

---

So what does 1,2, 3 as column header or the 13 mean in first row? I am really not sure what you are actually counting. Define "identical response per individual". Perhaps if you provided a small example such as a data step with 3 individuals, 4 questions and what the result matrix looks like.